Forged File Detection and Steganographic content Identification (FFDASCI) using Deep Learning Techniques

M. Srinivas, Akshay Nayak, Abhishek Bhatt
2019 Conference and Labs of the Evaluation Forum  
This paper presents our contribution in the identification and detection of Forged files and Steganographic content using Deep Neural Networks like Convolutional Neural Network and 3D-RESNET. We have used CNN in our research as CNN's are inspired by visual cortex. In other words, they are designed to extract consequential features which are relevant in classification i.e. the ones which minimizes the loss function. In this the kernel weights are learned by Gradient Descent so as to generate the
more » ... perceptive features from images fed to the network which in result supplemented to fully connected layer that performs the final classification task. In our proposed approach we mainly consider the two different tasks. Firstly, Identification of Forged Images has been carried out in which detection of altered images which includes both extension and signature has been performed. In addition to this, we have predicted the original epitome of forged file by using convolutional neural network model which automatically classify them and are useful for large-scale image classification as it has increased ConvNet depth. Secondly, we have recognized the Steganographic content by applying 3D-RESNET. Here, we have given preference to Residual Networks in place of VGG16 as increasing the depth should increase the accuracy of network, as long as over-fitting is taken care of. In VGG16 increased depth is increasing the effect of vanishing gradient and degradation problem. In this work, ImageCLEF 2019 data set is used for identification of Forged Images and recognized the Steganographic content.
dblp:conf/clef/SrinivasNB19 fatcat:j4nscgxvr5ao7fpdoqwdtsagne